[feat] Enhance PDF Text Extraction with Combined Embedded Text and OCR

Improve the PDF text extraction pipeline by leveraging both embedded text from PDFs and text generated via OCR.

The workflow should extract embedded text, process the PDF pages with OCR, and combine the outputs intelligently to produce an enriched result that retains structure (e.g., Markdown from OCR) and ensures accuracy (by validating against embedded text).

- [ ] Investigate libraries for extracting embedded text, such as PyPDF2
- [ ] Decide on the library to use and integrate it into the project
- [ ] Add utility function extract_embedded_text(pdf_bytes) to handle text extraction
- [ ] Use it in LM strategies combined with OCR results to improve the quality of the output - ask AI to check if there is something missing :)

Start by reviewing the existing PDF extraction pipeline and LM strategies, then investigate PyPDF2 as requested. Define how embedded text and OCR output should be combined and validated; done means the utility and strategy integration produce enriched, structured extraction results.

Written by the indexing model from the issue text.